05-26-2019, 06:11 PMaaronwilson Wrote:

05-26-2019, 05:17 PMFordyford Wrote: I'm afraid I feel obligated to point out that per your last post (with the descriptions of what the statistics mean), you stated that PDO is directly correlated to +/-. This may be true within a team (I'm yet to test that) but for the SHL as a whole, PDO to plus/minus has a PMCC of 0.08790, with n approx. 200, which wouldn't provide evidence for correlation at a 10% level, let alone the 5% I would typically use. (Inspecting it further, the value for the SMJHL does provide evidence for a correlation, although not by much. These statistics do not seem to be the strongest correlation, as one might expect)

that's probably because higher PDO doesn't mean higher +/-
rather, it helps to determine the sustainability of +/-
i.e. it's possible to have a good +/- and a low PDO

2 simple examples:

Player 1
10 goals for / 100 shots for = 0.100 shooting %
9 saves / 10 shots against = 0.900 save %
PDO = 1000 * (0.100 + 0.900) = 1000
Corsi = 100/110 = 91%
+9

Player 2
9 goals for / 10 shots for = 0.900 shooting %
10 saves / 10 shots against = 1.000 save %
PDO = 1000 * (0.900 + 1.000) = 1900
Corsi = 10/20 = 50%
+9

Same +/-, two completely opposite players.
Player 1 has a neutral PDO (1000) so it's sustainable. His +/- is so great because of good corsi and pace.
Player 2 has a very high PDO (1900) so his +/- is inflated by luck. Obviously, 0.9 sh% and 1.0 sv% aren't sustainable.

---

Or here's a real life example:

Borromini Cannellini: +14, 56.68 CF, 998 PDO
(oof) Aaron Wilson: +13, 55.94 CF, 1019 PDO

The same team, almost the same +/- and corsi but by looking at PDO you can suggest that I've been a little lucky this season.

There may be some value in doing a ranked correlation test (I don't have the ability to do that easily for a bit), But you are correct, in that 2 statistics with high variance which are also correlated may have very low correlations in some samples. These 2 statistics I imagine have large variance even for generally good players.